Sebastien Goddijn led internal AI at Ramp, where he built Glass, the platform that lets anyone at the company use AI, whatever their technical background. He was given a month to prove the idea, and within weeks half the company was using it every day. Ariella Taub sat down with him for our Building Glass webinar to talk through how he built it so fast, and what he believes are the fundamentals of AI tools people actually use.
Highlight Reel
Key Ideas from Sebastien
Spend the time defining context; it's the foundation of everything.
At Ramp, tiger teams across departments spent hours putting company wisdom on paper: documenting workflows and defining what good looks like. That knowledge becomes reusable company IP that AI can draw on from day one. Without it, the output is too generic.
You don't need to build something complex to create meaningful value. Focus on memory, skills and connectors.
Make sure each session carries useful context. Package company best practices so employees don't have to recreate them. Make it easy for AI to find the information it needs. Together, these make even a simple AI experience significantly more useful.
Build what's unique to you; buy what helps you build it faster.
Don't recreate infrastructure someone else already does well. Use vendors, SDKs and building blocks to move faster, but make sure whatever you buy gives you a foundation you can keep building on and customizing.
Be clear on ROI before you scale.
AI spend can grow very quickly without a clear view of the value it creates. Start small and identify a benefit you can actually track. Expand gradually, rather than opening the floodgates and trying to unwind expensive tools later.
Full Webinar with Sebastien Goddijn
The full conversation, including what Ramp learned once it had 1,500 skills, why you should try building something yourself before choosing a vendor, why open-weight models deserve more attention, and how Glass won users without a mandate.